The CITP Interconnection Measurement Project provides first-of-its-kind, direct measurement of interconnection capacity and utilization with a tool implemented by Internet service providers (ISPs). The visualizations linked below provide various views into the data.

Understanding the Data

The above views provide a more granular picture of the distribution of interconnect links across peak utilization levels. The heat map provides an hourly view of interconnect utilization across the 1200+ aggregated link aggregation groups (ALAGs) in the data set for a period of one month. Each horizontal row represents a single ALAG, and the height of the row indicates relative capacity of each ALAG. The shade represents its level of utilization. As illustrated in the heat map, the data shows a sizable proportion of capacity with low peak utilization levels. Diurnal patterns are also clearly observed for peak utilization levels.

The two histogram views provide complementary views of interconnects by peak utilization. The first, “distribution of interconnect links by peak utilization,” provides the number of ALAGs that fall within each 5-percentage point bin of peak utilization. Of the 1200+ ALAGs in the data set, only 2.1 percent reach a peak utilization of 95 percent or greater in November (note data outage for December 2016 that caused omission of histograms and heat map for that month, as discussed in more detail here). In comparison, nearly 78 percent (or, over 900 ALAGs) have a peak utilization of under 50 percent in November. The second, “distribution of interconnect capacity by peak utilization,” provides the portion of interconnect capacity that falls within each 5-percentage point bin of peak utilization. Of the total interconnect capacity in the data set, 2.5 percent of capacity reached a peak utilization of 95 percent or greater in November. In comparison, 64.1 percent of capacity had a peak utilization of under 50 percent in November. It is important to note that high utilization levels do not necessarily correspond with negative quality of experience (QoE). The exact relationship between utilization and QoE is an area of emerging research.

Since each ALAG is composed of a number of interconnection circuits, the number of routes available for Internet data traffic is much higher than represented here.